Re: Make all Perl warnings fatal

On 2023-08-22 Tu 00:05, Michael Paquier wrote:
> On Mon, Aug 21, 2023 at 11:51:24AM -0400, Andrew Dunstan wrote:
>> It's not really the same as -Werror, because many warnings can be generated
>> at runtime rather than compile-time.
>>
>> Still, I guess that might not matter too much since apart from plperl we
>> only use perl for building / testing.
> However, is it possible to trust the out-of-core perl modules posted
> on CPAN, assuming that these will never produce warnings? I've never
> seen any issues with IPC::Run in these last years, so perhaps that's
> OK in the long-run.

If we do find any such issues then warnings can be turned off locally.
We already do that in several places.

>> Regarding the dangers mentioned, I guess we can undo it if it proves a
>> nuisance.
> Yeah. I am wondering what the buildfarm would say with this change.
>
>> +1 to getting rid if the unnecessary call to getprotobyname().
> Looking around here..
> https://perldoc.perl.org/perlipc#Sockets%3A-Client%2FServer-Communication
>
> Hmm. Are you sure that this is OK even in the case where the TAP
> tests run on Windows without unix-domain socket support? The CI runs
> on Windows, but always with unix domain sockets around as far as I
> know.

The socket call in question is for a PF_INET socket, so this has nothing
at all to do with unix domain sockets. See the man page for socket() (2)
for an explanation of why 0 is ok in this case. (There's only one
protocol that matches the rest of the parameters).

cheers

andrew

--
Andrew Dunstan
EDB:https://www.enterprisedb.com